Skip to content

What your store needs to sell through AI agents

A six-level ladder, a directory of tools and protocols with their sources, and a plan per platform. No single score: you see which level you are at and what is missing.

8 questions · no sign-up · public rules

  • We list Trusteed's competitors
  • We label “Our product”
  • Every fact with a source and a date
  1. Level 1: DiscoverableCan an agent find my store and my products?
  2. Level 2: UnderstandableDoes it understand catalog, prices and policies?
  3. Level 3: OperableCan it check stock, create a cart and get a quote?
  4. Level 4: PayableCan it pay with a mandate from the user?
  5. Level 5: TrustworthyDo I know which agent it is, and do I apply rules?
  6. Level 6: VerifiableIs there auditable proof of what happened?

See it from the agent’s side

Almost four minutes following an AI shopping agent through a store that isn’t ready, then through one that is. The six levels, in motion.

3 min 46 s · English narration · English captions · Illustrative scenario with fictional data.

Read the transcript

My customer wants grey running shoes. Size 42. Under €120. I open the store's page. 403. I'm refused as automated traffic, and the next check needs a human. My customer authorized me to buy, but the store can't tell a legitimate agent from a bot. Without access, my journey ends here. Even if I get in, the data contradicts itself. One product, three prices: page, markup, checkout. I can't tell which one is valid. I guess, and the total passes my limit. One mistake, and my customer overpays. I can't trust the stock either. The page says in stock. The cart says zero. One name matches three products. I can't be sure which one to buy. Without a unique reference, I could buy the wrong product. Every error has a cost. The customer overpays, returns the item, or is left without a receipt. The store loses the sale, faces a dispute, and has nothing to audit. And there's a cost the merchant never sees. If I can't read the catalog, I have nothing to offer my customer. I look for another store. If I can read it but can't buy, the result is the same: a lost sale. My customer won't wait. The purchase closes at the store next door.

A store needs six capabilities, in order. The Advisor calls them levels.

If I can't find your store, it doesn't exist for me. A readable robots file, a sitemap and feeds: now you're on my shortlist. For the merchant: show up when I compare stores, or don't exist.

Three prices, three sources. Without structured data, I guess, or I skip you. With structured data, I get one price, one stock level and one policy. With consistent data, the agent compares your product instead of discarding it.

Reading isn't enough. This is where I need to act. Through an API, I check stock, build a cart, and request a quote. If the store lets me act, I get as far as the cart.

A card form is designed for humans. If the store accepts my customer's payment mandate, I pay within the limit they set. If it doesn't, the sale is lost at the last step.

To the store, every bot looks the same. It can't tell who I am. My identity is signed, and rules decide: allow, ask a human, or block. So the store doesn't have to choose between letting anyone in or closing the door.

An email is not proof. If something goes wrong, who can verify it? Now covered operations leave a signed receipt that can be verified. That proof helps resolve a dispute, depending on the provider's process that accepts it.

Same request. This time, the store is ready. One price. A confirmed total. Payment inside the mandate. My identity is checked, the human approves, and a signed receipt closes the purchase. Every step leaves proof. For the merchant, a completed sale. Which level is your store at? Find out with the Advisor. It's eight short questions, and it asks for no personal data.

How it works

  1. You answer 8 questions

    Your platform, your markets, your resources and where you are today at each level. No personal data and no store domain.

  2. We apply public rules

    We filter the tools that fit you and score them on compatibility, coverage, maturity and effort. The weights are published.

  3. You get a plan with the why

    For each level you are missing: the recommended tool, the reasons behind the choice, alternatives and a 30, 60 and 90-day work order.

The six levels

Each level answers a checkable question. They build on each other, but you can have one without the previous; the advisor flags that as a risk.

  1. Level 1

    Discoverable

    Can an agent find my store and my products?

    If an agent cannot find you, you do not exist for it. It is the foundation of everything else.

    How it is checked

    • robots.txt that does not block legitimate agents
    • sitemap
    • llms.txt
    • feeds
  2. Level 2

    Understandable

    Does it understand catalog, prices and policies?

    An agent decides with structured data: price, stock, returns. Without it, it guesses or skips you.

    How it is checked

    • Schema.org Product/Offer
    • MerchantReturnPolicy
    • ShippingDetails
    • GTIN identifiers
  3. Level 3

    Operable

    Can it check stock, create a cart and get a quote?

    Here the agent stops reading and starts acting: it checks stock, builds a cart and asks for a quote.

    How it is checked

    • UCP manifest
    • MCP or WebMCP endpoint
    • catalog via API
  4. Level 4

    Payable

    Can it pay with a mandate from the user?

    To get paid through an agent you need a mandate from the customer and a gateway that accepts it.

    How it is checked

    • AP2/ACP checkout
    • network tokenization
    • compatible payment gateway
  5. Level 5

    Trustworthy

    Do I know which agent it is, and do I apply rules?

    When bots buy, you need to know which one it is, what it may do and when to stop it.

    How it is checked

    • Agent identification (Web Bot Auth, KYA)
    • rules by trust level
    • anti-fraud
  6. Level 6

    Verifiable

    Is there auditable proof of what happened?

    If something goes wrong, signed proof of what happened helps in a dispute or an audit.

    How it is checked

    • Signed trust receipts
    • Public JWKS
    • proven human authorization (passkey, EUDI Wallet)
    • qualified seals

Plan by platform

What exists today at each level for your platform.

Tools

A tool is the concrete product you install or buy. Each profile explains what it does, what it gives you, what it costs and when it was verified.

Updated on · 42 tools · 11 protocols · How we verify

Platform
Price
Status
Covers

Showing 42 of 42 tools

Diagnostics

7 tools

What it doesThey measure how an agent sees your store today and what it is missing.

Why it mattersThe best place to start: it tells you which level you are at before you install anything.

Discovery and catalog

5 tools

What it doesThey publish your catalog in formats an agent can read: feeds and structured data.

Why it mattersIf an agent cannot find or understand your products, it will not buy from you.

Operation

14 tools

What it doesThey let an agent check stock, build a cart and ask for a quote in your store.

Why it mattersHere the agent stops reading and starts acting: without this stage it only sees you, it does not buy.

Payment

3 tools

What it doesThey let you get paid for a purchase made by an agent, under the customer's mandate.

Why it mattersWithout a payment the customer has authorized, the purchase does not close.

Trust

11 tools

What it doesThey identify which agent reaches your store and apply rules on what it may do.

Why it mattersWhen bots buy, you need to know which one it is, what it may do and when to stop it.

Verification

2 tools

What it doesThey leave signed proof of what happened that a third party can check.

Why it mattersIf something goes wrong, that proof helps in a dispute or an audit.

Protocols

A protocol is the shared language that lets an agent and a store understand each other. Picking tools that speak the same protocols keeps you from being stranded.

How we choose tools

Before it enters the directory, every profile goes through these tests. If it fails them, it stays out.

Selection criteria

  1. It exists and can be usedA product that is available, or a published specification. If access is by request, it enters as beta and we say so.
  2. We can read itPublic official documentation. If we cannot read it, we do not include it, however well known it is.
  3. It serves a merchant selling to agentsIt fits at least one of the six levels. Tools that are only for administration or development stay out.
  4. It says what it does, and only thatEach profile states coverage per level, status and verification date. Anything we could not confirm carries a note.
  5. It plays by the rulesNo practices that contradict the model, such as ignoring robots.txt or bypassing CAPTCHAs. Reviews and popularity do not score.

Honesty and neutrality

We are Trusteed: 8 of the 42 tools are ours and carry the «Our product» label.

Whether a tool is ours does not enter the score. An automated test checks it, and if it is the only option for your case, the plan says so.

We recommend competitors when they fit better. If there is no mature solution, we say so instead of filling the gap.

Nobody pays to be listed or to rank higher. If sponsorships ever exist, they will be marked and will not change the order.

What we cannot promise

  • We read the official documentation; we have not tested every tool.
  • It is a dated snapshot, not the whole market: something may be missing.
  • It is guidance, not legal advice or a certification.
  • The Advisor guides and the /audit audit measures what's visible from outside. Neither certifies anything. The technical evaluation happens with your store connected.
See the full methodology

EU regulation

eIDAS 2.0, the EUDI Wallet, PSD3/PSR and the AI Act applied to agentic commerce, with the official source for every date and obligation.

See the regulation

Methodology

How we decide what is listed and how we score, with the weights published.

Read the methodology
Agentic commerce advisor: what your store needs